Generalized Multi-Protocol Label Switching (GMPLS) is under development by the Internet Engineering Task Force (IETF) to provide a generalized control plane which can be used to control networks having a range of different network technologies, including packet-based networks and photonic networks.
One aspect of GMPLS is recovery of traffic. The term “recovery” includes protection of traffic and restoration of traffic. Recovery schemes are described in RFC 4426 (Generalized Multi-Protocol Label Switching Recovery Functional Specification). A shared meshed is one of the schemes most commonly deployed by network operators in which protection paths for multiple Label Switched Paths (LSPs) share common link and node resources. Under these schemes, the protection capacity is pre-reserved, i.e. link capacity is allocated to protect one or more LSPs, but explicit action is required to instantiate a specific protection LSP. This requires restoration signaling along the protection path.
RFC 4872 describes RSVP-TE Extensions in Support of End-to-End Generalized Multi-Protocol Label Switching Recovery. FIG. 1 shows an example network topology to explain how the recovery scheme works. There is a first working LSP [A-B-C-D] and a second working LSP [H-I-J-K]. A recovery path for the first working LSP is [A-E-F-G-D] and a recovery path for the second working LSP is [H-E-F-G-K]. The path [E-F-G] is shared by the two recovery paths. When a failure is detected on one of the working LSPs (e.g. on link B-C), the error is propagated to the ingress node (A). Ingress node (A) activates the protection LSP [A-E-F-G-D] for that working path. This means that the protection path [H-E-F-G-K] is not available. RFC 4872 describes that a node on the working path should notify the head node on the other working path that the resources for the protecting LSP are no longer available, but does not describe how this can be achieved.
A link-state routing protocol such as Open Shortest Path First (OSPF-TE) can propagate information about the resources of the shared path, following transfer of traffic to the protection path. However, the time period between the shared resources being used by one working path and the other working path finding out that the shared recovery path is no longer available can result in a heavy loss of traffic for the second working path. For example, in the case of a photonic network, a single LSP can carry traffic at up to 40 Gbs.
The present invention seeks to provide an alternative way to operate a network in which there is a shared recovery path.